Papers by Keyword: Computer Vision

Abstract: Detecting microscopic defects in precision manufacturing remains a major challenge, particularly in hard disk drive (HDD) production where sub-millimeter dust particles on the Voice Coil Motor Assembly (VCMA) can cause performance degradation or early device failure. This study presents a comparative evaluation of three YOLO object-detection architectures—YOLOv5, YOLOv8, and YOLOv11—applied to high-resolution dust detection on VCMA components. All models were trained and tested using the same annotated 5-megapixel dataset under identical experimental settings to ensure fair comparison. The results show that YOLOv5 achieved the highest precision (0.640) and the highest mAP50–95 (0.253), indicating stable localization performance across strict IoU thresholds. YOLOv8 produced the highest mAP50 (0.500), reflecting strong localization accuracy at IoU 0.5, while maintaining moderate precision (0.633) and lower recall (0.455). YOLOv11 obtained the highest recall (0.636), successfully capturing the largest proportion of true dust particles, though with lower precision (0.335) and weaker mAP values, revealing a higher rate of false detections. Overall, the findings highlight clear trade-offs among the models: YOLOv5 offers the most balanced performance, YOLOv8 excels in spatial localization, and YOLOv11 is suitable for scenarios where maximum defect coverage is prioritized. These insights support the selection of appropriate detection architectures for automated micro-defect inspection and contribute to the development of AI-driven quality-control systems in HDD manufacturing.

Abstract: This study explores the required Level of Detail (LOD) in 3D urban models to elicit observation responses similar to those in real spaces. Through experiments involving 30 participants, both real-world and 3D-modeled streetscapes were evaluated using psychological surveys and webcam-based eye-tracking. Results showed that higher model precision generally produced responses closer to those from real environments. However, inconsistencies appeared at higher LODs, likely due to fatigue or equipment limitations. Open horizontal spaces attracted greater attention, suggesting the need for detailed modeling in such zones. While a clear threshold of sufficient detail was not found, the findings highlight the potential of 3D models as substitutes for field observation and the necessity for standardized LOD criteria in urban simulations.

Abstract: This study presents the design and pilot evaluation of a computer vision–based system for monitoring exercise performance in older adults, aiming to reduce frailty-related risks without the need for wearable sensors. Using MediaPipe and OpenCV, the system tracks posture and movement in real time and provides feedback on exercise execution. A pilot test was conducted with 14 volunteers performing seven exercises from the Vivifrail Spanish program (Wheel A). Performance was evaluated using perfomance analysis, yielding recognition rates between 91.06\% and 100\% across exercises. While the system showed high accuracy in detecting posture and repetitions, challenges such as camera positioning, clothing variability, and the absence of validation in the target population remain. These findings demonstrate the feasibility of computer vision for exercise monitoring and support its potential as an accessible tool for fall prevention and functional assessment in older adults. Future work will focus on clinical validation and integration into mobile platforms for home based use. This approach will allow older population to adequately perform exercises from the Vivifrail program, while professionals, as physiotherapists and geriatricians, can monitor their progress remotely, adjusting the program when needed.

Abstract: Infant growth is an important indicator of their health and development, making accurate and regular measurements essential. However, traditional measurement methods are often inaccurate and require frequent visits to healthcare centers. In this report, we developed a infant height measurement application using smartphone camera technology and image processing algorithms to provide a practical and accurate solution. The application not only simplifies height measurement but also facilitates the recording and analysis of infant growth data. Parents can store measurement history digitally and share this data with doctors or healthcare professionals for a more comprehensive assessment. This feature is especially useful for the early detection of potential growth issues, such as stunting, allowing for timely medical intervention. Testing shows that the application has varying degrees of accuracy, with some subjects demonstrating high accuracy while others require further adjustments for consistent results.

Abstract: Traffic congestion has been a major problem in big cities, including Yogyakarta, with negative impacts including time, economic, and psychological losses. Based on data from the Yogyakarta Special Region Transportation Agency and Yogyakarta City Transportation Agency, analysis of congestion level data, and field observations, it was found that one of the main causes of congestion on the most congested roads in Yogyakarta City is vehicles parked on the side of the road. The proposed solution involves roadside parking detection and warning using surveillance cameras integrated with Artificial Intelligence (AI). The proposed system involves vehicle detection with pre-trained deep learning models, parking detection algorithms with Intersection over Union (IoU) tracking, and alerts that are forwarded to motorists as well as authorities such as the Transportation Department and local traffic police. The Yogyakarta CCTV dataset is used to test parking detection using various models, such as YOLOv5-medium, YOLOv5-large, YOLOv7-tiny, and Haar Cascade. The model evaluation shows that YOLOv5-large provides the highest accuracy of 86.1% with a processing speed of 5.5 Frames Per Second (FPS) to perform parking detection. With this proposed system, this research can contribute to solving congestion problems and improving traffic conditions in Yogyakarta City.

Abstract: Today's computational capacity enables the use of advanced statistical algorithms to identify relationships between features in high-dimensional data. Additive manufacturing methods are typically complex processes with many variables in both printing parameters and material properties. Consequently, machine learning offers opportunities for process optimization, quality assurance, and innovation in both Material Extrusion and Powder Bed Fusion technologies. The paper reviews the recent findings in machine learning applications for these additive manufacturing techniques, focusing on areas like defect detection, process control, and material property prediction. Key trends reveal that, while machine learning offers promising enhancements for additive manufacturing, challenges remain in data scarcity, model generalization, real-time adaptability. Our findings underscore the potential of machine learning to improve the overall quality of additive manufacturing processes by predicting optimal manufacturing parameters.

Abstract: In the recent era- very frequently people come across health issues due to consumption of poor-quality food items- which leads to issues such as food poisoning, vomiting, diarrhea, etc., For a full development of fruits and vegetables, all the nutrients are necessary during its growth. But due to circumstances like soil defects, infections, water scarcity, waterlogging, etc., the vegetables & fruits gets infected with some diseases. So there arises a necessity of a system which inspects for any presence of disease in fruits & vegetables, with reduced manual intervention. This paper provides a detailed overview of a system developed using the Python programming language. Its aim is to recognize and classify various fruits and vegetables, while also identifying any diseases affecting them and determining the specific type of infection. In order to recognize the details accurately, the system is designed to use convolutional neural networks (CNN) and the results are displayed using computer vision techniques. The analysis, implementation, and future improvements of the proposed system are briefed in this paper. For this, we have used Anaconda navigator software (Jupyter notebook, IDLE).

Abstract: This project focuses on developing and implementing a robotic control system based on detecting signs and gestures using computer vision. The main goal was to create an intuitive and efficient interface for interacting with an OMRON Viper 650 industrial robot. To achieve this, computer vision technologies like Mediapipe and OpenCV were used to detect and recognize the user’s hands and fingers in real-time. The collected data was processed with a Python script and stored in a text file. Additionally, a program was developed in C# using OMRON’s ACE programming interface to extract data from the text file and send commands to the Viper 650 robot, enabling it to interpret the user’s gestures and perform actions accordingly. This project has successfully created an innovative solution that combines computer vision, programming, and industrial robotics to provide an intuitive and efficient control experience, opening up new possibilities in industrial and human-robot interaction applications.

Abstract: Acoustic testing is a technology that covers various machinery failure modes, including bearing and gear failures. This technology is superior to vibration analysis for gear and bearing condition monitoring. This paper aims to offer the maintenance world a critical technological advance by developing a web-based tool that, using pretrained convolutional neural networks and spectrograms, allows the diagnosis of gearboxes from recordings obtained with industrial acoustic testing tools. The resulting model is tested against human specialists to assess its actual world performance. A modified agile methodology was implemented to develop the research systematically. The type of approach is mixed since it has qualitative parts, such as specialists involved in obtaining the ultrasonic data and classifying them, and quantitative parts, such as validating the precision of the model based on established validation metrics. By using a pretrained model and then performing a fine-tuning with heterodyne ultrasound recordings from gearboxes in good and bad condition, a training accuracy of 93% was achieved. Then, tests were carried out to validate false positives and negatives in which it was possible to obtain 0% and 6.7% scores, respectively. This model was incorporated on a web platform to create the diagnostic tool whose input variable is the recording, and the output variables are its spectrogram, the prediction of whether it is in good or bad condition, and the probability of both possibilities.

Abstract: The construction industry generates a significant amount of waste, posing challenges for efficient waste management and resource recovery. This paper presents a preliminary study on the use of lightweight computer vision (CV) algorithms for the automatic recognition of construction and demolition waste (CDW) materials. Utilizing image datasets acquired by drones, the study aims to develop strategies for distinguishing between individual CDW materials based on the mean intensity gradient, brightness, and relative representation of color channels. Results indicate that the proposed method can effectively recognize crucial CDW materials, paving the way for potential applications in industry and geodesy. Further research is needed to test additional materials and metrics to refine the method for practical implementation.